GISCafe
   >> GIS - Remote Sensing; Imagery
Thread views: 16261 View all threadsNext thread*Threaded Mode

mrasser
(Stranger )
04/20/08 05:08 PM
Manual Accuracy Assessment new Report this article as Inappropriate to us !!!Login to Reply

I have a set of accuracy data (400 pts)  that consists of a table with two columns.  One column represents the reference and the other the mapped value as shown below.  Does anyone know of a way to cross tabulate this data to get an error (confusion) matrix?  At this point I am considering doing it manually, but it seems like there should be a tool or database process that can accomplish this?

 

Any advice would be greatly appreciated!

 Ref  Mapped

1           1

3           2

3            3





mysticsea
(Stranger )
04/29/08 07:06 AM
Re: Manual Accuracy Assessment [re: mrasser]Report this article as Inappropriate to us !!!Login to Reply

You may try looking at this article from the University of Alaska.  Hope it helps.

http://nrm.salrm.uaf.edu/~dverbyla/nrm641/labs/classification_accuracy_lab_2008.pdf





wanakev2
(Stranger )
07/10/12 10:05 AM
Re: Manual Accuracy Assessment new [re: mrasser]Report this article as Inappropriate to us !!!Login to Reply

hi,


 


why dont you just calculate the RMS error? Laughing





scrabblehack
(Newbie)
07/11/12 06:14 AM
Re: Manual Accuracy Assessment new [re: wanakev2]Report this article as Inappropriate to us !!!Login to Reply

RMS error is good for determining positional error.  To determine classification error, you need a confusion matrix (you could use Cohen's kappa to reduce the matrix to a single number).


 





LColson
(Stranger )
07/12/12 06:54 AM
Re: Manual Accuracy Assessment new [re: mrasser]Report this article as Inappropriate to us !!!Login to Reply

If you have 400 points of accuracy data, your confusion matrix should have the 400 points spread between the classes. Example:


                  Urban        Non-urban


Urban           50            60


Non-Urban     90            200


Notice that the 4 numbers add up to 400. Only the diaganol is where an urban pixel in the reference data was also classified as urban (50) and your classification of non-urban areas was also matched with a non-urban reference pixel (200). 250/400 = overall accuracy. Do a search on producer's and user's accuracy to figure them out and then also Kappa Coefficient. But I would also recommend you look at techniques for comparing your classification results with your reference data to create a matrix where all the data in your results add up to 400.


The squared R approach is favored by many remote sensing analysts, but I think you are trying to do the approach I discussed above. I hope this helps. Smile        






View all threadsNext thread*Threaded Mode
Jump to

 

Trimble
CADalog.com - Countless CAD add-ons, plug-ins and more.



Internet Business Systems © 2016 Internet Business Systems, Inc.
595 Millich Dr., Suite 216, Campbell, CA 95008
+1 (408)-337-6870 — Contact Us, or visit our other sites:
AECCafe - Architectural Design and Engineering EDACafe - Electronic Design Automation TechJobsCafe - Technical Jobs and Resumes  MCADCafe - Mechanical Design and Engineering ShareCG - Share Computer Graphic (CG) Animation, 3D Art and 3D Models
  Privacy Policy